US Shut Down Claude Mythos Over China Data Fears, Not Jailbreak 2026

The US government's decision to force Anthropic to pull access to its Claude Mythos model was driven less by a specific jailbreak warning and more by growing suspicion that a foreign telecom with alleged China ties had access to the powerful AI, according to a Wired report. The order, issued last Saturday, required Anthropic to revoke access to both Fable 5 and Mythos for any foreign national, whether inside or outside the United States.

While the official justification centered on a vulnerability that could bypass the model's safeguards, sources close to the government told Wired that Washington was already uneasy about another issue: South Korea's SK Telecom, one of the major companies with access to Mythos, the original unsafeguarded model. The concern, according to these sources, stems from the operator's supposed links to China.

What Happened: The Order and the Official Story

Last Saturday, the White House ordered Anthropic to immediately restrict access to its advanced AI models, Fable 5 and Mythos, to any foreign person or entity. The move came after a warning that a method existed to jailbreak the model and bypass its safety guardrails, a claim that was reportedly flagged by Amazon CEO Andy Jassy to the Treasury Secretary.

Fable 5 was the public-facing, "trimmed-down" version of Mythos, while Mythos itself was the full, unsafeguarded model available only to select partners. The official narrative painted the decision as a rapid response to a technical vulnerability. But according to Wired, that warning arrived at a moment when the US government was already concerned about something else entirely.

The Real Concern: SK Telecom and China

According to sources cited by Wired, one of the major companies with access to Mythos was SK Telecom, South Korea's largest mobile operator. The US government's worry, these sources say, was not just about the jailbreak but about SK Telecom's alleged ties to China. The logic was blunt: if SK Telecom has access, then China has access.

That suspicion, combined with the vulnerability warning, created a perfect storm. As one source put it, the thinking was: "We've also been told there are vulnerabilities. Shut it down now." The decision was not merely a technical patch but a geopolitical move aimed at preventing potential foreign espionage or misuse of a powerful AI model.

Why This Matters: A Shift in AI Export Controls

Until now, the prevailing explanation for the shutdown was the jailbreak warning on Fable, the public version of Mythos. Amazon's CEO Andy Jassy reportedly called the Treasury Secretary to alert him that vulnerabilities had been found in Fable, which triggered the order. But the Wired report suggests a deeper motive: the US government is increasingly treating advanced AI models as strategic assets that must be protected from foreign adversaries, particularly China.

This incident sets a precedent for how the US might regulate AI exports in the future. It shows that access to cutting-edge models can be revoked not just for technical flaws but for geopolitical risk assessments. Companies with foreign ownership or ties to rival nations could find themselves cut off from the most powerful AI tools, even if they are based in allied countries like South Korea.

• Geopolitical risk trumps technical fixes: The US shutdown of Claude Mythos was driven by suspicions about SK Telecom's China ties, not just a jailbreak warning.
• Access is a privilege, not a right: Foreign nationals, even in allied nations, can lose access to advanced AI models if the US perceives a national security threat.
• Corporate ties matter: SK Telecom's alleged links to China were enough to trigger a sweeping order, highlighting how telecom operators are viewed as potential vectors for espionage.
• Precedent for AI export controls: This move signals that the US is willing to use export restrictions on AI models as a tool of foreign policy.
• Jailbreak warnings can be a catalyst: The vulnerability report on Fable provided the immediate trigger, but the underlying concern was already simmering.

Who Is Affected: Anthropic, SK Telecom, and Global AI Users

The order directly impacts Anthropic, which now faces the challenge of enforcing a nationality-based access restriction on its models. It also affects SK Telecom, which has not publicly commented on the allegations but now finds itself at the center of a US-China tech dispute. More broadly, any foreign company or individual that had access to Mythos or Fable 5 will lose it, potentially disrupting research, business operations, and AI development worldwide.

For the AI industry, the message is clear: the US government is willing to intervene directly in the deployment of advanced AI, even if it means overriding commercial agreements. This could lead to a fragmentation of the global AI ecosystem, with companies forced to choose between US-aligned and China-aligned technology stacks.

What to Watch Next: Escalation or Resolution?

The key question now is whether this order will be temporary or permanent. If the US government remains convinced that SK Telecom poses a risk, the restriction could become a long-term ban, prompting other countries to develop their own AI alternatives. Alternatively, if SK Telecom can prove its independence from China, access might be restored. The background to this story is set out in Claude Mythos AI Security: Hype vs. Reality in Cybersecurity. A related thread runs through Expiry date of ID cards issued as 27/06/2025 – The official announcement. Earlier coverage in Anthropic Faces Multibillion-Dollar Song Copyright Lawsuit shows how the argument developed.

Another factor to watch is how Anthropic responds. The company has not publicly detailed how it will enforce the nationality-based restriction, and it may face legal and operational challenges. Meanwhile, the incident is likely to accelerate efforts by other nations to build sovereign AI capabilities, reducing reliance on US models. A closer look at the same subject appears in Judge Rules Pentagon's Blacklisting of Anthropic Unlawful. Anyone following the wider picture can start with Iran Issues Tough Demands to Reopen Strait of Hormuz. The same theme surfaces in Anthropic Launches Claude Science for Laboratory Research, which takes a different angle.

The White House's order to Anthropic is a stark reminder that AI is now a matter of national security. As the US and China compete for technological dominance, access to advanced models like Claude Mythos will be scrutinized through a geopolitical lens, and companies with foreign ties may find themselves caught in the crossfire. Context on how this played out before sits in Anthropic: Claude May Aid Biological Weapons Development. A companion piece, Anthropic Settles AI Copyright Case, Bloomsbury to Receive, approaches the same subject from another angle.